Beyond Best Practices
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
Sebastian Bergmann argues that code reviews, comprehensive testing, static analysis, and documentation are not merely best practices but essential control mechanisms for safe agentic AI development. As AI agents shift from assistants to autonomous contributors, these practices define the specification layer that constrains AI behavior and preserves human agency. The piece covers TDD with AI agents (humans write tests, AI generates implementations), the role of ADRs and coding guidelines in constraining AI solution spaces, and the deskilling risk for junior developers who use AI heavily but benefit less than seniors. Bergmann also raises broader concerns about open source sustainability and platform dependency as agentic infrastructure concentrates in corporate hands.
Table of contents
From assistance to agencyAbstraction, not replacementCode Review DisciplineAI Safety NetTests eliminate ambiguityDocumentation: Knowledge and ConstraintsWhen AI is faster than learningPreserving human agencyThis is personalSort: